[PATCH 01/10] USB: OHCI: Properly handle ohci-at91 suspend
Alan Stern
stern at rowland.harvard.edu
Wed Jun 12 12:25:56 EDT 2013
On Wed, 12 Jun 2013, Alan Stern wrote:
> On Wed, 12 Jun 2013, Manjunath Goudar wrote:
>
> > Suspend scenario in case of ohci-at91 glue was not properly
> > handled as it was not suspending generic part of ohci controller.
> > Calling explicitly the ohci_suspend()routine in ohci_hcd_at91_drv_suspend()
> > will ensure proper handling of suspend scenario.
>
> > diff --git a/drivers/usb/host/ohci-at91.c b/drivers/usb/host/ohci-at91.c
> > index fb2f127..28400a7 100644
> > --- a/drivers/usb/host/ohci-at91.c
> > +++ b/drivers/usb/host/ohci-at91.c
> > @@ -619,8 +619,12 @@ ohci_hcd_at91_drv_suspend(struct platform_device *pdev, pm_message_t mesg)
> > {
> > struct usb_hcd *hcd = platform_get_drvdata(pdev);
> > struct ohci_hcd *ohci = hcd_to_ohci(hcd);
> > + bool do_wakeup = device_may_wakeup(&pdev->dev);
> > + int ret;
> >
> > - if (device_may_wakeup(&pdev->dev))
> > + ret = ohci_suspend(hcd, do_wakeup);
> > +
> > + if (do_wakeup)
> > enable_irq_wake(hcd->irq);
> >
> > /*
> > @@ -637,7 +641,7 @@ ohci_hcd_at91_drv_suspend(struct platform_device *pdev, pm_message_t mesg)
> > at91_stop_clock();
> > }
> >
> > - return 0;
> > + return ret;
>
> If ohci_suspend() fails, we should return right away. Don't execute
> the enable_irq_wake() and all the other stuff.
Actually, the situation is more complicated. In order to handle races
with remote wakeup, the logic we need is more like this:
Index: usb-3.10/drivers/usb/host/ohci-at91.c
===================================================================
--- usb-3.10.orig/drivers/usb/host/ohci-at91.c
+++ usb-3.10/drivers/usb/host/ohci-at91.c
@@ -674,8 +674,18 @@ ohci_hcd_at91_drv_suspend(struct platfor
{
struct usb_hcd *hcd = platform_get_drvdata(pdev);
struct ohci_hcd *ohci = hcd_to_ohci(hcd);
+ bool do_wakeup = device_may_wakeup(&pdev->dev);
+ int ret;
- if (device_may_wakeup(&pdev->dev))
+ ret = ohci_suspend(hcd, do_wakeup);
+ if (ret == 0 && do_wakeup && HCD_WAKEUP_PENDING(hcd)) {
+ ohci_resume(hcd, false);
+ ret = -EBUSY;
+ }
+ if (ret)
+ return ret;
+
+ if (do_wakeup)
enable_irq_wake(hcd->irq);
/*
@@ -692,7 +702,7 @@ ohci_hcd_at91_drv_suspend(struct platfor
at91_stop_clock();
}
- return 0;
+ return ret;
}
static int ohci_hcd_at91_drv_resume(struct platform_device *pdev)
The same is true for all the other drivers.
Alan Stern
More information about the linux-arm-kernel
mailing list